Publications

11-10-2006

Robust Control Techniques Enabling Duty Cycle Experiments Robust Control Techniques Enabling Duty Cycle Experiments

Utilizing a 6-DOF Crewstation Motion Base, a Full Scale Combat Utilizing a 6-DOF Crewstation Motion Base, a Full Scale Combat

Hybrid Electric Power System, and Long Distance Internet Hybrid Electric Power System, and Long Distance Internet

Communications Communications

Marc Compere Science Applications International Corporation, comperem@erau.edu

Jarrett Goodell Science Applications International Corporation, jarrett.d.goodell@saic.com

Miguel Simon Science Applications International Corporation

Wilford Smith Science Applications International Corporation

Mark Brudnak U.S. Army RDECOM-TARDEC, brudnakm@tacom.army.mil

Follow this and additional works at: https://commons.erau.edu/publication

Part of the Hardware Systems Commons, Military Vehicles Commons, Propulsion and Power

Commons, and the Software Engineering Commons

Scholarly Commons Citation Scholarly Commons Citation Compere, M., Goodell, J., Simon, M., Smith, W., & Brudnak, M. (2006). Robust Control Techniques Enabling Duty Cycle Experiments Utilizing a 6-DOF Crewstation Motion Base, a Full Scale Combat Hybrid Electric Power System, and Long Distance Internet Communications. , (). Retrieved from https://commons.erau.edu/publication/710

This Report is brought to you for free and open access by Scholarly Commons. It has been accepted for inclusion in Publications by an authorized administrator of Scholarly Commons. For more information, please contact commons@erau.edu.

\CQCo~~

2006-01-3077

Robust Control Techniques Enabling Duty Cycle ExperimentsUtilizing a 6-DOF Crewstation Motion Base, a Full Scale

Combat Hybrid Electric Power System, and Long DistanceInternet Communications

Marc Compere, Jarrett Goodell, Miguel Simon, Wilford SmithScience Applications International Corporation

Mark Brudnaku.s. Army TARDEC

ABSTRACT

The RemoteLink effort supports the U.S. Army'sobjective for developing and fielding next generationhybrid-electric combat vehicles. It is a distributed soldier­in-the-Ioop and hardware-in-the-Ioop environment with a6-DOF motion base for operator realism, a full-scalecombat hybrid electric power system, and an operationalcontext provided by OneSAF. The driver/gunnercrewstations rest on one of two 6-DOF motion bases atthe U.S. Army TARDEC Simulation Laboratory (TSL).The hybrid power system is located 2,450 miles away atthe TARDEC Power and Energy System IntegrationLaboratory (P&E SIL). The primary technical challengein the RemoteLink is to operate both laboratoriestogether in real time, coupled over the Internet, togenerate a realistic power system duty cycle.

A topology has been chosen such that the laboratorieshave real hardware interacting with simulatedcomponents at both locations to guarantee local closedloop stability. This layout is robust to Internetcommunication failures and ensures the long distancenetwork delay does not enter the local feedback loops.The TSL states and P&E SIL states will diverge due to(1) significant communications delays and (2)unavoidable differences between the TSL's power­system simulation and the P&E SIL's real hardware-in­the-loop power system. Tightly coupled, bi-directionalinteractions exist among the various distributedsimulations and software- and hardware-in-the-Ioopcomponents representing the driver, gunner, vehicle, andpower system. These interactions necessitate additionaladjustment to ensure that the respective states at theTSL and P&E SIL sites converge. This is called stateconvergence and ensures the dominant energetic statesof both laboratories remain closely matched in real time.State convergence must be performed at both locationsto achieve bi-directional, real-time interaction like that

found on a real vehicle. The result is a distributedcontrol system architecture with Internet communicationsin the state convergence feedback loop.

The Internet communication channel is a primary sourceof uncertainty that impacts the overall state convergenceperformance and stability. Multiple control schemes weredeveloped and tested in simulation. This paper presentsrobust control techniques that compensate forasynchronous Internet communication delays duringclosed loop operation of the TSL and P&E SIL sites. Thesubsequent soldier- and hardware-in-the-Ioopexperiments were performed using a combination ofnonlinear Sliding-mode and linear PID control laws toachieve state convergence at both locations. The controlsystem development, performance, and duty cycleresults are presented in this paper.

Report Documentation Page Form ApprovedOMB No. 0704-0188

Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering andmaintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information,including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, ArlingtonVA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if itdoes not display a currently valid OMB control number.

1. REPORT DATE 10 NOV 2006

2. REPORT TYPE N/A

3. DATES COVERED -

4. TITLE AND SUBTITLE Robust Control Techniques Enabling Duty Cycle Experiments Utilizinga 6-DOF Crewstation Motion Base, a Full Scale Combat HybridElectric Power System, and Long Distance Internet Communication

5a. CONTRACT NUMBER

5b. GRANT NUMBER

5c. PROGRAM ELEMENT NUMBER

6. AUTHOR(S) Marc Compere; Jarrett Goodell; Miguel Simon; Wilfor Smith; Mark Brudnak

5d. PROJECT NUMBER

5e. TASK NUMBER

5f. WORK UNIT NUMBER

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) SAIC 8303 N. MoPac Expy. Suite B-45 Austin, TX 48759, USE USArmy RDECOM-TARDEC 6501 E 11 Mile Rd Warren, MI 48397-5000

8. PERFORMING ORGANIZATION REPORT NUMBER

9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) US Army RDECOM-TARDEC 6501 E 11 Mile Rd Warren, MI48397-5000

10. SPONSOR/MONITOR’S ACRONYM(S) TACOM/TARDEC

11. SPONSOR/MONITOR’S REPORT NUMBER(S) 16688

12. DISTRIBUTION/AVAILABILITY STATEMENT Approved for public release, distribution unlimited

13. SUPPLEMENTARY NOTES Presented at SAE Power Systems Conference, November 2006, New Orleans, LA, USA, Session: AdvancedPower Systems Technologies

14. ABSTRACT

15. SUBJECT TERMS

16. SECURITY CLASSIFICATION OF: 17. LIMITATIONOF ABSTRACT

SAR

18. NUMBEROF PAGES

10

19a. NAME OFRESPONSIBLE PERSON

a. REPORT unclassified

b. ABSTRACT unclassified

c. THIS PAGE unclassified

Standard Form 298 (Rev. 8-98) Prescribed by ANSI Std Z39-18

Introduction

A series of Duty Cycle Experiments (DCE) have used thegrowing fidelity of TARDEC's Power Budget Model andRemoteLink technology to gain increasingly useful dutycycles for designing the Army's Future Combat System(FCS) family of vehicles. Duty Cycle Experiments 1.0and 1.1 were simulation-based experiments with realtime soldier-in-the-Ioop inputs and physics-based models[1,2]. This paper presents the RemoteLink controlsystem developed for Duty Cycle Experiments 2.0 and2.1. The 2.x experiments extend the 1.x experimentsand incorporate a full scale combat hybrid electric powersystem at TARDEC's Power and Energy SystemIntegration Laboratory (P&E SIL) in San Jose California.A standard Internet connection was used to connect thetwo laboratories in real-time coordinated experiments2,450 miles apart (Figure 1).

TSL, Warren MI

Figure 1: TARDEC Power and Energy Lab and TARDECSimulation Lab

The goal of Duty Cycle Experiment 2.x was to captureoperational duty cycles using Army soldiers in a realisticvirtual environment with appropriate Visual, sound, andmotion cues while interacting with a full-scale combathybrid electric power system. A military vehicle's dutycycle is specific to the mission and platform type but isa design- and configuration-independent representationof events and circumstances which affect powerconsumption. Such events andcircumstances encompass (1) vehicle operation alongthe course such as speed, grade, turning, turret/gunactivity, and gun firing plus (2) external scenariocomponents that affect power consumption like incomingrounds, ambient temperature, and soil conditions. Theevent inputs can be distance based when the vehicle ismoving or time based when the vehicle is stationary, oreven triggered with some other state condition. Withinthe scenario, the duty cycle inputs result in a use-historyspecific to that particular vehicle design and componentchoice. The use-history outputs may then be used tocompute the vehicle's performance and cost metricsassociated with the duty cycle.

System and Experiment Description

The Warren, Michigan based TARDEC Simulation Lab(TSL) assets include a family of 6-DOF Stewart platformmotion bases and driver/gunner CAT crewstations[1,2,12]. The Ride Motion Simulator (RMS) and TurretMotion Base System (TMBS) are hydraulically-actuatedStewart platforms designed to provide realistic motioncues to the operators driving in a virtual environment.The position, velocity, and acceleration inputs to themotion bases came from a high-fidelity tracked vehiclemodel implemented in SimCreator [7]. The vehiclemodel interacted in real time with CHPSPerf, which is aMatlab/Simulink based hybrid electric power systemmodel [10]. Two crewstations provided realistic driverand gunner visuals via touch-screen interfaces and ayoke/pedal combination. Whether driver or gunner, theappropriate touch-screen interface was displayed alongwith a high-resolution 3D rendering of the surroundingsand appropriate audio cues. The surroundings includethe terrain, roads, trees, buildings, power lines, and otherfriendly or enemy vehicles. The TSL used OneSAF TestBed (OTB) to provide the operational context in whichthe soldiers perform their missions [13].

The P&E SIL houses a full scale combat hybrid electricpower system in a highly instrumented laboratoryenvironment [3]. The objective power system was aseries hybrid with a 250kW diesel engine/generator, two410kW traction motors, and a 50 kW-hr battery packconnected via a 600V bus [4]. Over 120 sensors wererecorded to capture the power system's duty cycleperformance. Mobility loads were imposed in the labusing bi-directional dynamometers coupled to a local realtime tracked vehicle model [4]. Non-mobility loads wereimposed on the power system using a 250kWAerovironment AV-900 bi-directional power supply [6].For DCE 2.x, the power system under test was similar tothe objective power system except a single tractionmotor was operational rather than two. To achieverealistic power system results the second traction motorwas simulated in software and the associated mobilityload or supply was imposed on the hardware using theAV-900.

Neither the TSL nor the P&E SIL are portable, so toachieve the real-time bi-directional interaction betweenthe driver/gunner and power system hardware, anInternet-enabled configuration was chosen to couple thetwo labs with the soldier-in-the-Ioop and hardware-in-the­loop (Figure 2).

Figure 2: System Layout (top boxes in TSL, Michigan;bottom two in P&E SIL, California)

Due to scheduling and technical constraints the DCE 2.0experiments achieved uni-directional communicationfrom the TSL to the SIL with a series of trained Armytracked vehicle drivers and gunners. This effectivelymade the SIL a TSL follower and made the TSL's powersystem response solely a function of the CHPSPerfpower system model. The DCE2.1 experiments usedengineers as driver and gunner subjects with bi­directional Internet communications and achieved bothmobility and power system state convergence at bothlocations,

To manage the divergence, Mobility State Convergencewas implemented to facilitate coordinated laboratoryoperation while keeping both mobility models at thesame location in their respective virtual environments.Similarly, Power System State Convergence was used toensure the TSL's power system model followed the P&ESIL's relevant power system states. Both stateconvergence implementations used 30Hz updates over astandard Internet connection. A simulation study wasperformed to evaluate the feasibility of RemoteLinkoperation, characterize the Internet communication, anddevelop control strategies [11].

The experiment included human presence at both theTSL and P&E SIL facilities where the 6-dof motion baseplatform and the power system HWIL were located,respectively. As such, safety was of the utmost concern,first with regard to human safety and secondarily withregard to equipment protection. As stand alone devices,both the 6-DOF motion base simulator and the P&E SILpossess robust, multi-layer safety interlock and faultdetection systems which are intended to detect andprevent hazardous situations from developing. However,by coupling the systems together, the possibility of newhazards are introduced which are unknown to the stand­alone systems. These additional hazards arise fromthree potential sources (1) the unreliability of the chosencommunication channel, the Internet, (2) the stability ofthe closed loop system, and (3) the divergence of thestate convergence algorithm.

Internet communications delay and drop ratecharacteristics between the TSL and SIL werebenchmarked. Results indicated that the packet lossrate was approximately 0.1 % and the most likely roundtrip time was 94ms with some jitter in the network delay.Since the loss rate was so low, we chose to accept thatsome packets would be lost and chose UDP/IP as ourcommunication protocol over TCP/IP. The benefit ofTCP's retry feature was not worth the risk of a 1 (s) ormore delay during the retry attempt. At approximately30Hz, losing one UDP packet only cost 0.03(s) withanother packet right behind. As SUCh, our design did notassume a particular update rate, instead each controlaction was tied to a particular event, that being the arrivalof a new packet. This approach has been shown robustin the presence of the unreliable and multi-path nature ofthe Internet [14]. Additionally, to guard against Internetconnection loss the packet delay and update intervalwere continuously monitored and reported to the safetysubsystem.

Finally, state convergence algorithm divergence wasmonitored continuously and compared to presetthresholds to assure the errors remained bounded. Inthe event the threshold was violated, the SIL would drop

Safety Considerations

To address the stability of the system, the control actionsin the state convergence were intentionally designed todrive the states to their reference values over a period oftime which is greater than the network delay. In this way,the control action is not likely to over correct before thearrival of updated information. Delay in a closed loopsystem is a notorious source of instability. Normally thisis caused by error computation using data which aretime-skewed by the delay. By, using an event-basedapproach similar to that described in [14], error can becomputed using time-synchronized data thus alleviatingdelay-based instabilities. The system was too complexto rigorously prove stability. Therefore we testedextensively prior to running the experiment with people orhardware in the loop to obtain confidence that the systemwould be stable.

Vehicle Dynamics andTerrain

S roc etTorques

Hybrid Power Train (SIL)

d~~,~">Vehicle Dynamics and

Terrain Observer

Motion

•---- -- -_.---- --.a..UI ._'1Power Train Observer

Throttle,Steer,Brake

Throttle,Steer,Brake

Driver (TSL)•

Identical 3D high-fidelity tracked vehicle mobility modelswere implemented at both locations. The TSL's mobilitymodel interacted in real-time with the Ride MotionSimulator and also the Power Train Observer,CHPSPerf. Driver and gunner commands were sent tothe simulated power system, CHPSPerf, whichdeveloped simulated traction motor torques. The motortorques independently spun the left and right sprocketswhich drove the TSL's mobility model over digitizedterrain, Simultaneously, the driver and gunner inputswere sent over the Internet to the P&E SIL's powersystem hardware which, in turn, developed torques anddrove the SIL's local mobility model. Because bothlocations contained separate mobility models, themethods for achieving closed-loop stability at eitherlocation remained localized and independent of theInternet communication's delay and jitter. However,using two mobility models opened the possibility that thetwo would diverge because of differences betweensimulated and real power system responses.

out of the experiment and the TSL would continue instand-alone mode. Also, automated safeguards wereimplemented where the TSL and SIL both sent statusand health signals to the other location. In the event thata health signal became false, the RemoteLinkconnection would be broken, the SIL shut down, and theTSL experiment would continue in stand alone mode.The TSL monitored three aggregate P&E SIL statusmessages: (1) vehicle dynamics status, (2) powersystem hardware status, and (3) mobility stateconvergence status. Likewise, the P&E SIL alsomonitored three TSL aggregate status signals: (1)vehicle dynamics status, (2) power system stateconvergence status, and (3) motion base hardwarestatus.

Control System DesignConsiderations

Vehicle velocity convergence is achieved if position timehistories match, but the opposite is not true. Also, thetrack-terrain interaction induces sprocket and motorspeed convergence when both vehicle positions trackwell. During preliminary simulation-based testing it wasfound that the sprocket speed error changed withfrequency higher than the network update rate.However, the global position and yaw error signalschanged at low enough frequency such that the 30 HzInternet update rate provided good signal resolution.Network update rate was the most important factor formobility state convergence. Because the TSL driverthrottle, steer, and vehicle position all arrived at the SILin the same time stamped UDP packet the stability ofmobility state convergence was independent of the one­way delay. Mobility state convergence error wascomputed in the SIL's delayed time frame,

emobility SC (t +~ 1 ) =Ydes (t) - Yact (t +~ I ) .

Figure 3: Mobility State Convergence Ensures BothModels Remain Close

The RemoteLink technical objectives for the 2.x series ofDuty Cycle Experiments are listed below:

For mobility state convergence, the global XY positionand vehicle heading ( yaw) were determined sufficient tocoordinate both vehicle models.

• The RemoteLink should maintain realistic driverperception and feel - this means the motionbase should ensure realistic driver/gunner inputs(no video game inputs) plus the TSL's drivershould perceive a torque response similar to theP&E SIL's power system hardware

• The RemoteLink should facilitate meaningfulpower system results - this means the P&ESIL's power system hardware should providerealistic performance for a combat hybrid electricvehicle including realistic tracked vehicle mobilityloads.

• The RemoteLink should ensure state tracking atboth sites - this means both mobility modelpositions and velocities should remaincoordinated and the TSL's power system modelshould follow the real hardware's bus voltage,both in the presence of variable Internetcommunication delays.

The error was also used in the delayed time frame. Thismeans the driver inputs, vehicle states, and SIL stateadjustments were consistent with each other,independent of the delay, and simply caused the SIL'svehicle model to lag behind the TSL by the one-waydelay, ~1'

e power system SC (t + ~l ) =Vbus,SIL (t + ~1 ) - Vbus,TSL (t)

This is the reason power system state convergencestability was influenced more by the round trip delay thanmobility state convergence. Power system stateconvergence implemented changes in the TSL's presenttime with errors computed ~2 seconds in the past. Also,that error was computed using a TSL bus voltage

e power s)'stem SC (t + ~1 +~2) =Vbus,SIL (t +~1) - Vbus,TSL (t)

Neglecting SIL computation time, the bus voltage errorcomputation was performed in the SIL and, althoughdelayed, represents an error consistent with the timehistory of the SIL and TSL. However, to be used in theTSL, the error experienced the second network delay, ~2'

Using absolute times, the power system stateconvergence adjustment used in the TSL was at t + ~1 +~2 and a function of error computed with two different,although consistent, bus voltages,

The objective set of states for Power System StateConvergence included bus voltage, battery state ofcharge, engine speed and multiple coolant temperatures.To meet schedule constraints scope was reduced toinclude bus voltage for DCE 2.x. The TSL's simulatedbus voltage was sent to the SIL with the same UDPpacket containing the driver inputs and vehicle position.The bus voltage error was then computed in the SIL inthe time frame delayed by ~1, which was tSIL = tTSL + ~1'

Using absolute times, the error was

YlSIL Follower~

~Xl"',"'------- Yl

,,: Xl

III

: TSL LeaderIIII

I X

y

sampled at Ll1 + Ll2 seconds in the past. Despite delayinfluencing the error, stability was maintained with bothmobility and power system state convergence.

into the body-fixed 6-DOF equations of motion, Plongitudinal

and P'atera' along with one angular acceleration, Pyaw.

Driver inputs

Mobility State Convergence

The (gscp) term represents the mobility stateconvergence inputs. The f(x) represents the nominalplant with control inputs, (g·u). The plant and (g·u)control inputs are discussed below after mobility stateconvergence inputs and outputs are defined.

The final set of outputs were X position, Y position, andheading in the inertial, or terrain frame, lfI. These outputshad no observability problems since both vehicletrajectories remained unique and were generated insimulation. Other possible outputs considered were leftand right sprocket speed, vertical position in the terraindatabase, vehicle pitch and roll, and xyz vehicle velocity.Due to network update rate reasons discussed above,the sprocket speeds were removed from the outputvector. Also, the vertical terrain database position andpitch and roll angles were removed from the set ofcontrol outputs because, assuming vehicle positions,headings, and terrain databases were coordinated, thelocal vehicle model dynamics and terrain interactionwould produce similar vehicle states without explicitcontrol action.

Mobility state convergence is intended to ensure bothvehicle mobility models remain within some maximumdistance of each other within the virtual environments.Position and heading agreement were particularlyimportant while traversing bridges and negotiating otherobstacles that would have generated significantlydifferent terrain response in the TSL versus the SIL.Since the TSL was man-rated and operated with asoldier-in-the-Ioop, all mobility model state adjustmentswere made in the SIL. The TSL vehicle was the mobilityleader and the SIL model the follower. Focusing on theSIL vehicle, the mobility state convergence problem is anonlinear MIMO system with variable time delay, threeinputs, and three outputs. The time delay is not modeledexplicitly, therefore the nonlinear system in the SIL maybe represented with,

State ConvergenceAlgorithm

.1 Find p-fen(e,x)1~I Suchthal e~O I

[

X] [glObal X PositiOn]y == Y == global Y position

'¥ vehicle heading

I )' =TSL vehicle statesx= f(x)+ guy~h(x)

GVSL Power SystemModel and Vehicle Mlldel•

[P&ESILJ

. I @=1-'p == mohility state L----t ~ == f(~) +g. U +gsc' pL-.3"

(0rnm"",, input )' = h(x) S' = SIL vehicle states

SIt Power Systemand Vehicle Model

1/ ::0< throttll',hrau,steer,gear >

The second input topology previously evaluated usedthrottle and steering inputs to achieve mobility stateconvergence. This topology was associated with the H­infinity methodology which remained a viable option,however was not pursued for these experiments. Theadvantage of using throttle and steering inputs weresimplicity and straightforward interpretation andimplementation. The risk with these inputs is loss ofcontrollability. Either power system saturation or high­slip track-terrain conditions could cause the controlinputs to no longer affect the outputs, at least for sometime. Neither of these conditions presented themselvesfor extended periods during normal Duty CycleExperiment operation.

These fictitious accelerations provided controllabilityguarantees in the presence of power system and track­terrain saturation. No limitations in the track-terrainmodel or power system model would prevent theskyhook inputs from achieving state convergence. Also,skyhook inputs retained clear physical meaning byentering the equations of motion as acceleration inputsand provided gradual position and velocity adjustment.Velocity level inputs would have caused instantaneousmomentum changes and as a result would have riskedmultiple shutdown conditions while interacting with thepower system hardware. Abrupt vehicle speed changestranslate into abrupt motor speed setpoint changeswhich could cause over-current shutdown conditions.

Figure 4: Mobility State Convergence Control SystemDiagram

(1)x = f (x) + g .u + g . psc

In addition, several design choices were made to reducethe set of mobility SC inputs from several down to three.In the simulation study documented previously [11], bothH-infinity and Sliding mode control methodologies wereevaluated along with two input vector topologies. Thefirst input topology associated with Sliding mode usedthree fictitious skyhook inputs to achieve mobility stateconvergence. Two linear accelerations were inserted The inputs from (3) and outputs from (2) define the plant

and nominal inputs in (1). In the SIL they contain both

The final choice of mobility SC control inputs was acombination of the two input topologies previouslystudied. The output vector was composed of lateralskyhook acceleration, Plalerah yaw skyhook angularacceleration, Pyaw, and the driver's augmented throttleinput, thsc,

.v=[X Y If/Y (2)

p = [P'areral P yaw thscr (3)

simulated and real hardware states. The plant'spresentation as !(x)+ g·u with additional mobilitystate convergence inputs is an abstraction thatencompasses multiple subsystems and closed loopcontrol systems within. For example, because thethrottle augmentation is an input in (3), the f (x) + g .uin (1) represents not only the 6-DOF vehicle model'sequations of motion, but also the engine/generatorcomponents and controllers, the traction motors withtheir current and speed controllers, the battery and itscontrol, the bus voltage and control, the dynos andmobility load emulation control, and the driver'sthrottle/steer/gear inputs.

Power System State Convergence

facilitating a 42 minute duty cycle experiment with realtime bi-directional signal flow between both laboratories.

Mobility Results

The XV plot below is a top-view with both TSL and SILpositions overlaid. Mobility state convergence achievedan average of approximately 5(m) error or less with oneexcursion of 30(m). The excursion event resulted froman unexplained -7(s) TSL computer freeze. This isindicated by the spike in the network delay figure below.There is no discernable difference in vehicle positions onthe plot below because of scale. The course wasapproximately 8 miles long and took approximately 40minutes to drive.

TSL and SIL ,",hicle positions o,",oaid

10L-_----l_-L_~ __.L-_--'--_ __"____~_~_~

35 4 ~5 55 6 55 7 75XPos (km)

finish::0 I I

12

18

:: I131OS;::=:I==~=_=_=_J;::j_=_=_::J_=;::::=::J;-_0-_.Tjr=_=_Lt-==_~~

____ r :__ - __ ~ _ sta~ ; 1I I I ~_

I I I I13.102 - - - - I- - - 1- - - - - I- - I

E I I I I I

e I I I I I I:g 16 - - -lf3 - - - I - - - I - '""'

"- I 3.98 I 4 I 4.'02 I 4.04' I I>- I I I I I I I I

I I I I I I I

14 - -~r--:--_:L - --i- ---\ ---1- --r--V'I I I I I I i

--~---T---r---I---~---T---r--

I I I I I

I I I I

I I I I

The goal of power system state convergence was tocause the relevant states of the TSL power systemmodel, CHPSPerf, to track the relevant states in the realSIL power system. The power system stateconvergence problem was also a nonlinear system withvariable time delay and reduced to a SISO system. A PIcontroller was selected to achieve reasonable busvoltage tracking for DCE 2.1. The controller used busvoltage error to apply an artificial current to the highvoltage bus model in the TSL. This artificial currentdrives the TSL's bus voltage to SIL's hardware voltage.The gains in the PI controller were tuned in order to meettwo goals: 1) TSL bus voltage tracking the P&E SIL busvoltage 2) The driver in the TSL shouldn't notice anysudden vehicle oscillations due to the artificial currentsupplied by the powertrain observer.

Driver Inputs and Mobility State Con\€rgence Periormance

Figure 5: Power System State Convergence ControlSystem Diagram

II =< rl1rorrfe.hrakl' ..I!ff'r.Xt'ar >

y = Vhus.CHPSpeif

Ifind P - fenCe, x) ISuch thaI e --) 0 I

Sf e=Y-5'

1----'

Y =""JUs,SILSIL Puwer System

.i=fU)+g·u+p

v= h(i)

SIL Power Model

Sensor DataL-. From SIL

Hardware

Driver inpul~

I

•

Duty Cycle Experiment Results

This section presents DCE2 experiment results focusingon state convergence performance and influence on bothlaboratories. The results show both mobility stateconvergence and power system state convergence

Measured Intemet Delay Time7,-- --- ---,---- -------,-------,------,---------c

30002500

500

1000 1500 2000Time (s)

~ S4°c.a.~M_~ilNtwil~g, Ii15 I> - - - - _I I' _ _

I I, I

II I

I I

480 - - - - -: - - - - - - -: - - -- - - -

I I,

460 - - - - -, - - - - - -

600

580

620i--,----,-----,------r,::===:;iISIL \oS TSL bus I.Qltage

560

300025001500 2000Time (s)

I

I__1 1.- __. _

I I

I I

I I____ I --<- - _-i -l- __ - _

I I I

I I II I I

- - - - - - -i - - - - - - + - - - - - -I I

I I

I I- - - - -I - - - - 1 - - - - ~ - T - - - - - -

I I

I II ,

-'1-------------II I

- :- ~_=_:.:.c:~~J:....:J...:_:L:~....- - - - +~I I I I

, I I

'I1L - - - -I '-

oL-__----'-----_500 1000

Power System Results

The figures above indicate the TSL and P&E SILvehicles stayed within a few meters of each otherthrough the entire RemoteLink experiment. Exceptionswere few and resulted from computer operating systemunresponsiveness which caused large communicationdelays.

300025001000 1500 2000GVSL sim time (s)

500

I

I,

____I --t_,I

I

II I I I I

- - - - -I _; .... ~ 1 -' - -I- __ - -

I I I I I

I I I I I

I I I I I

I I I I I

I I 1 I I- - - - -I - - - r - - - - - 1- - - - - ..., - - - - - ,- - - - --

"'"-E~ -50

_aJ

-100

-150

0

The product of the bus voltage and artificial current, 'sc, isthe artificial power that is supplied to the CHPSPerfpower system by the powertrain observer. Theintegration of the artificial power over the time duration ofthe Fort Knox run results in a quantity called "leakedenergy", which is discussed below.

~ I II, I I

50 - - - - -:- - - -'r ~ - - - - i:- ----~ ----~ ---I'" ,I I I

I I I I

I I

I

Power System State Con",rgence Current Input

o

The bus voltage errors stay in a reasonable range for theduration of the run. Similarly, the artificial power systemstate convergence current input stays within 100 ampswith one exception.

300025001500 2000Time (s)

Mobility State Con",rgence Inputs

1000

The plots below show the bus voltage and artificialcurrent inputs for a representative DCE 2.1 experiment.

Leaked Energy

Leaked energy is calculated for both the powertrainobserver running in the TSL and for the vehicle dynamicsobserver running in the P&E SIL. As described above,power system leaked energy results from the integrationof artificial power over the length of the experiment. Thisintegration results in units of Joules for leaked energy. Inthe vehicle dynamics observer, the leaked energy iscalculated from the product of skyhook accelerations,masses, and velocities integrated over t,he length of therun. These quantities are normalized in the powertrainobserver and in the vehicle dynamics observer in orderto obtain percent leaked energy. In the vehicle dynamics

The vehicle's speed versus distance varies from stoppedto a top speed of 25mph. Cross-referencing the speedswith the main gun events and pulsed power eventsindicates the vehicle was both stopped and driving whileengaging the OPFOR SAF entities.

Experimental results from a 42 minute, 8 mile drive withthe TSL and P&E SIL communicating over the Internetwith bi-directional RemoteLink signals are presentedbelow. These figures are plotted with distance as theindependent axis. This allows direct comparison to othervehicles with different architectures or components.

8

3000

7

2500

63 4 5Distance (miles)

Velocity'" Distance

1500 2000Time (5)

Total AV900 Power", Time

2

1000

A Representative Duty Cycle

Non-Mobility Loads

Leaked energy percent is a metric that characterizeshow effectively the model approximates the equivalenthardware while operating under realistic real-timeconditions. In the case of the TSL power system model,a small leaked energy value of 0.6% indicates that themodel closely matches the behavior of the realhardware. Likewise, a 2% leaked energy value for themobility model in the P&E SIL indicates that the driverand gunner inputs were realistically implemented.

As mentioned above, the DCE's SIL power systemdiffered from the objective power system by simulatingone traction motor rather than implementing both tractionmotors in hardware. The simulated traction motor andsensor provided inputs to the SIL mobility model's virtualdriving environment but did not implement the real loadson the hardware power system. To emulate thesimulated motor's power draw in hardware, thecomputed power was added to the AV-900's non-mobilityload power for hardware interaction with the real powersystem.

The AV900 is a bi-directional power supply that can pushor pUll up to 125 kW of power to or from the bus. Theseloads include hotel loads, PFN power draws, auto-loaderpower draw corresponding to the main gun fire, andactive protection system power draws. The plot belowshows the total AV900 power draw and from the highvoltage bus during a representative DCE 2.1 experiment.

observer, leaked energy is normalized by the total energyinput to the vehicle dynamics model, which includes themechanical energy coming from the left and right motorsand the energy provided by the artificial skyhookacceleration inputs to the vehicle model. In thepowertrain observer, the leaked energy is alsonormalized by the total energy input to the high voltagebus, which includes the input energy from the battery andthe input energy from the generator. For DCE2 Run 1,the vehicle dynamics leaked energy was 0.6 percent andthe power system leaked energy was 2 percent.

Grade 1.5 Distance

15f~=~-----=:- i - -!- .10 r -- ~ -- ~ ---' c-

:~ ~ ~ , - ~ :~'"~,h.,ICc -j.b'

Q.l I I I~ -10 - -1- - - - I - I

" I I I-15 - - - _1- ~ _ _ _ _ __ J __

, ,

I I I I I

:::~- ~t-~~J~-~~:,-~~~!~~~~[~~~ ~~~r~~, , ,

-30 ~=,_:- - ~ - - - :-: - - - -:- - -=-~-",--'=---=--::.JL:-=--=-=--=--'l'-='--='-~-c:Jl

134 567Distance (miles)

Cross referencing the speed figure with the grade versusdistance figure, it appears the driver stopped on a 30%grade, most likely to scan a horizon or the other side ofthe hill, then moved forward and engaged the SAFenemy vehicles.

other SAF entities, and vehicle models interacting with afull-scale combat hybrid electric power system in realtime. The resulting electrical and mechanical duty cyclescan be used to further iterate designs for the Army'sfuture vehicle systems and components. Future DutyCycle Experiment plans include improved power systemhardware, improved pulsed-power loads, improvedmobility load emulation, improved round-trip networkdelay, and improved SAF and scenario interaction.

A control scheme was developed to couple twodynamical systems over the Internet and was shown toperform adequately in the presence of stochastic delayand packet loss. The control scheme adopted an event­based approach triggered by the arrival of newinformation. In this way, it was not necessary to estimateor compensate for the jitter in the network delay or theeffects of packet loss.

References

Conclusions

The RemoteLink architecture facilitated duty cycleexperiments in Summer 2006 using bi-directional real­time communications over the open Internet using Armyassets at laboratories 2,450 miles apart. The duty cycleexperiments featured live soldier-in-the-Ioop inputs with6-DOF motion and sound feedback, CAT crewstationdriver and gunner interfaces, 3D scenario images with

Pulse Forming Network (PFN) events happened in rapidsuccession at the 6.3 mile distance which is right afterthe driver stopped to scout out the area on a 30% hill.The PFN hits were imposed on the SIL's power systemwith a 50kW power draw for 4 seconds using the AV­900. The total PFN energy cost was 200kJ. Main gunfiring events occurred at roughly 4 and 6.5 miles into thecourse while engaging enemy SAF forces. Powersystem draws associated with these events were 1kWfor 4 seconds and implemented in the SIL using the AV­900.

1. Brudnak, M.; Nunez, P.; Paul, V.; Mortsfield, T.;Shvartsman, A.; Perera, H.S.; Smith, W.;Mohammad, S.; Pozolo, M., "Human-in-the-IoopSimulation-based Combat Vehicle Duty CycleMeasurement: Duty Cycle Experiment 1", PaperSIW-06S-080, Simulation InteroperabilityStandards Organization (SISO), Orlando, FL,April 2006.

2. Paul, V.; Brudnak, M.; Ueda, J.; Shvartsman, A.,"Simulation-based Hybrid Electric CombatVehicle Duty Cycle Measurement", IntelligentVehicle Systems Symposium, NDIA, TraverseCity, MI, June 2006.

3. Miguel Simon, Marc Compere, ThomasConnolly, Charles Lars, Wilford Smith, MarkBrudnak, "Hybrid Electric Power and EnergyLaboratory Hardware-in-the-Loop and VehicleModel Implementation," SAE 2006 WorldCongress & Exhibition, Military Vehicle Modelingand Simulation session, April 2006, Detroit, MI,USA

4. John Kajs, Marc Compere, Eugene Danielson,"Power System Upgrade for TARDEC SystemIntegration Lab," presented at The SixthInternational All Electric Combat VehicleConference 2005 (AECV 2005), Bath, England,June 13-16, 2005

5. Marc Compere, Miguel Simon, John Kajs, MikePozolo, "Tracked Vehicle Mobility LoadEmulation for a Combat Hybrid Electric PowerSystem", presented at The Sixth International AllElectric Combat Vehicle Conference 2005(AECV 2005), Bath, England, June 13-16, 2005

6. AV-900 Installation, Operation, and MaintenanceManual, AeroVironment Inc, Monrovia, California

7. Realtime technologies, Royal Oak, MI,www.Simcreator.com

8. McCullough, M. K. and Haug E. J., "Dynamics ofHigh Mobility Track Vehicles," ASME Design

87

----t---;, I

-,---,, ,-,------,-, ,

-i-----j-

- ..! - - - ---.!., ,

43

1F·~-- ---t-- --1----1----1-----(--

(/) - I I I I IC 0.8 - - T - - -I - - - -1- - - - r - - - T -

~ 0.6 - - ~ - - - - - - - -II - - -1- - - - ~ - - - ~ -

~ 0.4 - - ~ - - I _ _ _ ~ +_~ 0.2 I: - ~ - - - +-

o--------"-_-------"----._1 234 5

Distance (miles)

Engineering Technical Conference, Cincinnati,Ohio, Paper No. 85-DET-95, 1985.

9. Wong, J.Y., Theory of Ground Vehicles, 3rdEd.,John Wiley & Sons, ISBN 0-471-35461-9,2001.

10. Wilford Smith and Patrick Nunez, "Power andEnergy Computational Models for the Designand . Simulation of Hybrid-Electric CombatVehicles", SPIE Paper No. 5805-02, PaperPresented at Simulation Science IX, Orlando,FL, 28 March - 1 April 2005.

11. Jarrett Goodell, Marc Compere, Miguel Simon,Wilford Smith, Ronnie Wright, Mark Brudnak,"Robust Control Techniques for State Trackingin the Presence of Variable Time Delays," SAE2006 World Congress & Exhibition, MilitaryVehicle Modeling and Simulation session, April2006, Detroit, MI, USA

12. Truong, N.; Paul, V.; Shvartsman, A""Integration of the CAT Crewstation with the RideMotion Simulator (RMS)" SAE 2006 WorldCongress & Exhibition, Paper 2006-01-1171 ,April 2006, Detroit, MI, USA

13. OneSAF Website, http://www.onesaf.org.14. I. Elhajj, N. Xi, W.K. Fung, Y.H. Liu, Y.

Hasegawa, T. Fukuda, "Supermedia EnhancedInternet Based Telerobotics", in Proceedings ofthe IEEE, special issue on Networked IntelligentRobots Through the Internet, Vol. 91, No.3, pp.396-421 , March 2003.

CONTACTS

SAIC authors may be contacted at:

Dr. Marc Compere: CompereM@saic.comMr. Jarrett Goodell: GoodeIlJa@saic.comMr. Miguel Simon: SimonM@saic.comMr. Wilford Smith: SmithW@saic.com

Science Applications International Corporation8303 N. MoPac Expy. Suite B-450Austin TX 78759(512) 366-7812

The U.S. Army TARDEC author may be contacted at:

Dr. Mark Brudnak: brudnakm@tacom.army.mil

U.S. Army Tank Automotive Research Developmentand Engineering Center6501 E. 11 Mile Rd.AMSRD-TAR-N/157Warren, MI 48397-5000(586) 574-7355

DEFINITIONS, ACRONYMS, ABBREVIATIONS

DeE: Duty Cycle Experiment

OneSAF: A composable computer generated force thatcan represent a full range of operations, systems, andcontrol processes from entity to brigade level withvariable fidelity supporting multiple Army modeling andsimulation domain applications.

OPFOR: Opposing or opposition forces within a SAFenvironment

OTB: OneSAF Test Bed

P&E SIL: Power and Energy SIL

SAF: Semi Automated Forces

SIL: System Integration Laboratory

TSL: TARDEC Simulation Laboratory